|8th May 2017
For some people, being faced with mountains of data can be dauntingly boring. For others, it’s a stimulating challenge that cannot go unconquered. To make spreadsheets of numbers actually mean something, you need a mixture of both. Turning data into a meaningful and engaging story requires both creative and analytical thinking, and this is the exactly the approach we took in this year’s entry to the BHBIA Analyst Team of the Year competition. Now, it’s easy to confess that data analysis certainly isn’t the sexiest topic for an article, so instead of going into the details of how data analysis works, we’ve outlined the multi-disciplinary approach we used in our journey to the finals of the competition.
Analysing data and building a story is impossible without context, so the first step in any analytical project is to build a solid background of the therapy area (TA) in question. Context empowers you to ask questions you wouldn’t know to ask if you didn’t have a rock solid understanding in the TA.
Imagine you were tasked with identifying which regions in the UK a pharmaceutical company should focus their resources on to improve sales of a drug in their stroke portfolio. As well as the typical prevalence, competitor sales, and market value forecast data you’d expect to use, it’s also crucial to build a broader understanding of the TA:
Answering these questions will improve your ability to accurately and creatively break down the problem at hand.
A common problem-solving technique used within consulting is the use of mutually exclusive, collectively exhausted (MECE) issue trees (originally developed my McKinsey). In simple terms, it’s a framework for segmenting a problem into its component parts in a way that:
removes the potential for any overlap between components
Breaking down the problem using MECE will ensure awareness of the assumptions that confine your thinking and allow you to systematically address possible routes of the problems. In the profitability example below, if we systematically analyse each of the component parts, then we can be comfortable knowing that we have addressed the root of the problem.
When working with data, the MECE approach can be used as a starting point to develop initial hypotheses and identify any gaps in the data that is currently available.
A creative way of supplementing these hypotheses is to plot out the types of data that are available in a grid. In the example below, five different types of data are available and have been plotted against each other to identify potential analyses that could be relevant to the problem at hand. At this stage, it’s important to be creative in combining data types and question every assumption when carrying out this process. By doing so, you can infer innovative and creative hypotheses. In addition to combining data types together in this way, you can further ‘layer’ your hypotheses by adding additional variables to them.
As you map variables against each other, it’s critical to ensure that you’re keeping raw volumes relative to another metric. For example, 5% sales growth for product X may seem like a great achievement, but if sales across the industry grew 25% and the number of competitors stayed the same, then the company has fallen behind relative to its competitors.
After breaking down the problem using the MECE framework and then refining hypotheses by plotting variables in a grid format, data analysis can then occur to test the hypotheses and identify the messages you want to communicate. To build these into a story, it’s important to define the key takeaway for the reader, and then simply work backwards to identify the order of the messages that build up to the key takeaway.
We can then use the key messages from our analysis to identify the most appropriate visualisation methods. To do this, it helps to list out all the types of variables and identifiers (identifiers are the ‘codes’ or ‘names’ associated with a data point that be used to link data across different data sources) that together form the hypothesis. By doing so, you can brainstorm the various methods to communicate the components of the message that has come out of your analysis. For inspiration, it’s helpful to review public pieces of data analysis work on common analysis and visualisation tools, such as Tableau or Google Fusion Tables. Finally, after mapping the order of the messages and the types of visualisation for each message, you can then identify the most appropriate visualisation tools or software.
I was recently asked if, given the opportunity, whether the team would participate again, and my response was if we could do enter the competition again tomorrow morning, we would jump at the opportunity. At BLH, I’m more than proud to admit that we have a mix of people who find spreadsheets dauntingly boring and people who find them an exhilarating challenge. It’s this combination that propelled us to the finals of the BHBIA Analyst Team of the Year. With a multi-disciplinary team, we were able to combine the best of creative and analytic skills to break down millions (quite literally) of data points into a short, interactive, and meaningful story.
The four steps above, while broad, can be applied across any number of data analysis and visualisation projects. If you’d like to hear more about our approach to data analysis and visualisation, get in touch with Director Simon Young at firstname.lastname@example.org.
|25th April 2019
Patient advocate Trishna Bharadia explains why shared decision making is crucial for ensuring multiple sclerosis patients feel heard and empowered
|18th April 2019
David Lazarus charts his journey from initial MS diagnosis in 1990 to his experience participating in innovative clinical trials, along with his advice for other patients.
|10th April 2019
A new MIT research project, sponsored by Novo Nordisk, is aiming to deliver insulin orally with a pill that releases medicine in the stomach lining. Dina Patel interviews the team and reveals the innovative engineering behind the pill.